Measured adolescent body mass index and adult breast cancer in a cohort of 951,480 women.

Body mass index (BMI) in adolescence, studied predominantly as a self-reported risk factor for breast cancer (BC), may have been subjected to recall bias. We examined the association between measured BMI in adolescence and the incidence of BC by menopausal status. 951,480 Jewish Israeli females aged 16-19 who underwent anthropometric measurements in 1967-2011 were followed up to 31.12.2012 for BC incidence. Cox proportional hazards models assessed the association between adolescent BMI (as age-specific CDC percentiles) and time to BC diagnosis, adjusting for sociodemographic covariates. The analysis was also subdivided by age at diagnosis. 9619 BC cases diagnosed during 18,078,941 person-years of follow-up were included in multivariable analyses: 4901 premenopausal, 3809 perimenopausal, and 909 postmenopausal. Compared with 'healthy' BMI (5th-<85th percentiles) and adjusted for country of origin, education, and height, adolescent BMI was largely negatively associated with BC: hazard ratio (HR) = 1.057 (95 % confidence interval (CI) 0.955-1.169, p = 0.286) in underweight (<5th percentile); HR = 0.918 (95 % CI 0.849-0.993, p = 0.032) in overweight (85th-<95th percentiles); and HR = 0.682 (95 % CI 0.552-0.843, p = 0.00004) in obese (≥95th percentile) women. In premenopausal, but not peri- and postmenopausal BC, associations were statistically significant; underweight was associated with increased risk of premenopausal BC (HR = 1.15, 95 % CI 1.01-1.31, p = 0.037), and overweight and obesity with significantly lower risk. Adolescent thinness was associated with increased risk for early BC. Overweight and obesity were protectively associated with premenopausal but not postmenopausal BC. The lack of an association of adolescent overweight/obesity with increased peri- and postmenopausal BC suggests a causal role for adult weight gain.